Page 1 of 1

How to cause the voice to sound more like a question?

PostPosted: Wed Sep 26, 2007 5:39 pm
by troybowman
I'm using Allison for the voice. I wonder how I can make the end of the sentence gradually go up in pitch, make it sound like a question. Is there a way to do that? It seems to ignore question marks and ends a sentence going down in pitch. For example,

Code: Select all
$ swift "are you still there?"

sounds like "are you still there."

I tried to see what it'd be like with emphasis:

Code: Select all
$ swift "are you still <emphasis level='strong'>there?</emphasis>"

But that just makes "there" last longer. I can see how this would emphasize words, and I'd probably use it in conjunction with whatever else is out there to make the pitch gradually increase...

Anyway, I can't find the answer on the website, and I'm hoping it's possible.


PostPosted: Wed Oct 24, 2007 10:03 am
by jleoni
Review the SSML FAQ page and I think you'll find all of the answers you need ... &type=ssml

FAQ does not have this info

PostPosted: Wed Oct 24, 2007 10:20 am
by troybowman
Nope, sorry. This topic is not covered by that FAQ. That's where I got the emphasis idea, but it did not help. A question gradually raises in pitch. I can't artificially do that with the SSML pitch control, because the pitch gradually goes down no matter what pitch I tell it to start at. For example, try this:

Code: Select all
swift "are you still <prosody pitch='+50%'>there?</prosody>"

Nope, doesn't work. The pitch in the voice still goes down in the word. I want it to gradually increase the pitch in the word "there". I can't do that with the info given on the FAQ page.

PostPosted: Tue Jan 26, 2010 2:17 am
by dieter
Currently I am using this little trick to get kind of what you mean:

replace "?" by "<prosody volume='0'> 3 </prosody>?"

Compare eg. Linda, saying:

Do you know what time it is ?
I don't know.


Do you know what time it is <prosody volume='0'> 3 </prosody>?
I don't know.

The length of the silenced phoneme (here "3") can be varied by using other content, but as far as I see you can't get rid of the short break after the tags.

Re: How to cause the voice to sound more like a question?

PostPosted: Tue May 28, 2013 12:25 pm
by siyanz
I have a similar question and I have been using what dieter suggested. But it is not very satisfactory. I am trying to generate audio files for words with a question tone (a rising tone). I tried to embed the words into a sentence such as "What is a blah?" and then use Audacity to cut off the last part. But I can't get Allison to have a rising tone. :? Any suggestions?

Re: How to cause the voice to sound more like a question?

PostPosted: Tue May 28, 2013 3:02 pm
by NicholeH
I just emailed you a response, but I will paste it below as well. One thing I have to add is that "What is a blah?" is actually an information question, so the pitch should be falling towards the end of the sentenced, not rising. Here's a fun little youtube video that discusses the topic:

The best option for both version 5 and version 6 is still to use the SSML pitch tag as follows:

"Do you want to go to the <prosody pitch='+10st'>game</prosody>"

or even:

" <prosody pitch='+4st'>Do you want to go to the</prosody> <prosody pitch='+12st'>game</prosody>"

The other thing to keep in mind is that yes/no questions generally have rising intonation, whereas information questions tend to have falling intonation, relative to the beginning of the sentence. This means you might be able to get away with not changing the pitch for information questions.

Usually when I use SSML pitch tags, I play around with different pitch levels before I decide on one that works for the voice and the text being synthesized. Here is the link to our SSML tutorial, in case you'd like to browse through the options: